From Machine to Machine: An OCT-trained Deep Learning Algorithm for Objective Quantification of Glaucomatous Damage in Fundus Photographs

Purpose: Previous approaches using deep learning algorithms to classify glaucomatous damage on fundus photographs have been limited by the requirement for human labeling of a reference training set. We propose a new approach using quantitative spectral-domain optical coherence tomography (SDOCT) data to train a deep learning algorithm to quantify glaucomatous structural damage on optic disc photographs. Design: Cross-sectional study Participants: 32,820 pairs of optic disc photos and SDOCT retinal nerve fiber layer (RNFL) scans from 2,312 eyes of 1,198 subjects. Methods: The sample was randomly divided into validation plus training (80%) and test (20%) sets, with randomization performed at the patient level. A deep learning convolutional neural network was trained to assess optic disc photographs and predict SDOCT average RNFL thickness. Main Outcome Measures: The performance of the deep learning algorithm was evaluated in the test sample by evaluating correlation and agreement between the predictions and actual SDOCT measurements. We also assessed the ability to discriminate eyes with glaucomatous visual field loss from healthy eyes with the area under the receiver operating characteristic curve (ROC). Results: The mean prediction of average RNFL thickness from all 6,292 optic disc photos in the test set was 83.3 ± 14.5 μm, whereas the mean average RNFL thickness from all corresponding SDOCT scans was 82.5 ± 16.8 μm (P = 0.164). There was a very strong correlation between predicted and observed RNFL thickness values (Pearson’s r = 0.832; R2 = 69.3%; P<0.001), with mean absolute error (MAE) of the predictions of 7.39 μm. The areas under the ROC curves for discriminating glaucomatous from healthy eyes with the deep learning predictions and actual SDOCT average RNFL thickness measurements were 0.944 (95% CI: 0.912– 0.966) and 0.940 (95% CI: 0.902 – 0.966), respectively (P = 0.724). Conclusion: We introduced a novel deep learning approach to assess fundus photographs and provide quantitative information about the amount of neural damage that can be used to diagnose and stage glaucoma. In addition, training neural networks to objectively predict SDOCT data represents a new approach that overcomes limitations of human labeling and could be useful in other areas of ophthalmology.